Hi Isaac,
Thanks for the info but for the decomposed case the format you mentioned doesn’t work. So, the name of my case is
LEE_simple_source.
My case folder has 6 .chk folders and, and each folder contains 10 decomposed .fld files corresponding to “mpirun -np 10” parallel run. It also contains
an xml folder with, again, 10 xml files corresponding to “mpirun -np 10” parallel run. The parent case folder looks like this:
I tried to run the following command but it threw a bunch of segmentation faults:
mpirun -np 10 FieldConvert LEE_simple_source.xml LEE_simple_source_0.chk/ LEE_simple_source_0.vtu
Just to clarify, I am able to use FieldConvert for serial runs. No issues at all.
Another variation I have tried is this:
mpirun -np 10 FieldConvert LEE_simple_source_xml/P0000000.xml LEE_simple_source_0.chk/P0000000.fld LEE_simple_source_0.vtu
This in fact generates .vtu and .pvtu files as mentioned in the user guide. However, for some reason when I load .pvtu files, ParaView shows only
partial domain. On further investigation I found that the individual processor files are the same. Shouldn’t they be different? If you have an “mpirun -np 10” solution, each file should have unique data which when merged should show the whole domain.
I can send you my setup and mesh files if you have time. It’s nothing special. It’s a 50 x 50 2D quad mesh with a source placed at the centre.
Regards,
Ishan
PGR
From: Isaac Rosin <isaac.rosin1@ucalgary.ca>
Sent: 30 September 2023 20:13
To: Nande, Ishan A. <ian9@leicester.ac.uk>; nektar-users@imperial.ac.uk
Subject: Re: Using FieldConvert for parallel solution files
|
You don't often get email from
isaac.rosin1@ucalgary.ca.
Learn why this is important |
|
***CAUTION:*** This email was sent from an
EXTERNAL source. Think before clicking links or opening attachments. |
Hello Ishan,
Reconstructing a parallel solution does not need to be done in parallel. Perhaps with some very large solution files this is necessary, but in my experience, reconstructing in serial can be quite
efficient too.
Because you say your solution is decomposed, case_0.chk must be a directory. What I have done in the past is: FieldConvert case.xml case_0.chk/ case_0.vtu
The only differences between this and the command you were using are that I'm running in serial and I added a "/" to the end of the case_0.chk. You could try this with mpirun -np 10 in front as well. Just make sure you have the "/". If
you are using FieldConvert with a different number of processors than you ran the simulation with, this may not work but give it a try. If you still can't get it to work, please send your error message.
I would also recommend using the HDF5 output option which will tell your simulation to output solution files that are already reconstructed. You will still have to convert the files to a format readable by Paraview but it's helpful to have
fewer files in case you need to save many time steps and your cluster has a file count limit. There are several other benefits of HDF5 with Nektar++.
Please let me know if this helps and if you have more questions.
All the best,
Isaac